Recommendation in Education Portal by Relation Based Importance Ranking

摘要

Recommendation in education portal is helpful for students to know the important learning resources in schools. Currently, previous methods which have been proposed to solve this problem mainly focus on page view counts. A learning resource is important just because many students have viewed it. However, as the metadata in a resource is becoming available, the relations among the resources and other entities in real world are becoming more and more. Unfortunately, how to use such relations to make better recommendations has not been well studied. In this paper, we present a complementary study to this problem. Specially, we focus on a general education portal, which consists of different typed objects, including resource, category, tag, user and department. The recommendation object is resource. However, we have found that a resource's importance rank can be affected by its relations to other typed objects. Thus, we formalize the resource recommendation as a ranking problem by considering its relations to other typed objects. A random walk algorithm to estimate the importance of each object in the education portal is proposed. Finally, the experimental result is evaluated in a real world data set.